Modelling Bounded Rationality In Agent-Based Simulations using the Evolution of Mental Models

Abstract

A technique for modelling economic agents with bounded and procedural rationality is presented. In this technique an agent has a population of mental models which evolve. This evolution is its learning process. This is embedded within an agent architecture which then exhibits several qualitative characteristics relevant to modelling human-like agents. Two example models using this technique are presented. The first is of agents attempting to learn their own utility function in the presence of structural change. The second is an extension of Brian Arthur's `El Farol Bar' model where evolutionary learning and communication is added.